EPA/1 OO/K-09/001 I March 2009
                                                  www.epa.gov/osa
United States
Environmental Protection
Agency
                The U.S.  Environmental Protection
                Agency's  Strategic  Plan for
                Evaluating  the Toxicity of Chemicals
                         Chemicals
      aoiieiinuais  i—i

	ii  •  ^^ [  i  j^_«  I "*•'"-1 ['••••' I I
                                              Receptors / Enzymes / etc.
                                              Direct Molecular Interaction
                                              Pathway Regulation /
                                              Genomics
                                          Cellular Processes
                                     Tissue / Organ / Organism Tox Endpoint
Office of the Science Advisor
Science Policy Council

-------
                                                  EPA 100/K-09/001
                                                        March 2009
The U.S. Environmental Protection Agency's
          Strategic Plan for Evaluating
            the Toxicity of Chemicals
                 Office of the Science Advisor
                   Science Policy Council
             U.S. Environmental Protection Agency
                   Washington, DC 20460
                     Recycled/Recyclable
                     Printed with vegetable-based ink on paper that
                     contains a minimum of 50% post-consumer
                     fiber and is processed chlorine free.

-------
                                    DISCLAIMER
Mention of trade names or commercial products does not constitute endorsement or
recommendation for use. Notwithstanding any use of mandatory language such as "must" and
"require" in this document with regard to or to reflect scientific practices, this document does not
and should not be construed to create any legal rights or requirements.
                                          11

-------
                         AUTHORS AND CONTRIBUTORS


                   Future of Toxicity Testing Workgroup Co-Chairs
Michael Firestone, Office of Children's Health Protection and Environmental Education, U.S.
      EPA
Robert Kavlock, Office of Research and Development, U.S. EPA
Hal Zenick, Office of Research and Development, U.S. EPA

                             Science Policy Council Staff
Melissa Kramer, Office of the Science Advisor, U.S. EPA

                 Future of Toxicity Testing Workgroup Representatives
Marcia Bailey, Region 10, U.S. EPA
Arden Calvert, Office of the Chief Financial Officer, U.S. EPA
Laurel Celeste, Office of General Counsel, U.S. EPA
Vicki Dellarco, Office of Prevention, Pesticides, and Toxic Substances, U.S. EPA
Scott Jenkins, Office of Air and Radiation, U.S. EPA
Gregory Miller, Office of Policy, Economics, and Innovation, U.S. EPA
Nicole Paquette, Office of Environmental Information, U.S. EPA
Santhini Ramasamy, Office of Water, U.S. EPA
William Sette, Office of Solid Waste and Emergency Response, U.S. EPA

                                 Other Contributors
Katherine Anitole, Office of Prevention, Pesticides, and Toxic Substances, U.S. EPA
Hugh Barton, Office of Research and Development, U.S. EPA
Norman Birchfield, Office of the Science Advisor, U.S. EPA
Michael Brody, Office of the Chief Financial Officer, U.S. EPA
Rory Conolly, Office of Research and Development, U.S. EPA
David Dix, Office of Research and Development, U.S. EPA
Stephen Edwards, Office of Research and Development, U.S. EPA
Andrew Geller, Office of Research and Development, U.S. EPA
Karen Hamernik, Office of Prevention, Pesticides, and Toxic Substances, U.S. EPA
Jean Holmes, Office of Prevention, Pesticides, and Toxic Substances, U.S. EPA
Richard  Judson, Office of Research and Development, U.S.  EPA
Thomas Knudsen, Office of Research and Development, U.S. EPA
Julian Preston, Office of Research and Development, U.S. EPA
Kathleen Raffaele, Office of the Science Advisor, U.S. EPA
Ram Ramabhadran, Office of Research and Development, U.S. EPA
James Samet, Office of Research and Development, U.S. EPA
Patricia Schmieder, Office of Research and Development, U.S. EPA
Banalata Sen, Office of Prevention, Pesticides, and Toxic Substances, U.S. EPA
Imran Shah, Office of Research and Development, U.S. EPA
Linda Sheldon, Office of Research and Development, U.S. EPA
                                         in

-------
John Vandenberg, Office of Research and Development, U.S. EPA
Maurice Zeeman, Office of Prevention, Pesticides, and Toxic Substances, U.S. EPA

                               External Peer Reviewers
John R. Bucher, Ph.D., Associate Director, National Toxicology Program, National Institute of
      Environmental Health Sciences
George Daston, Ph.D., Research Fellow, P&G
Daniel Krewski, Ph.D., MHA, Professor and Director, McLaughlin Centre for Population Health
      Risk Assessment, University of Ottawa
Martin Stephens, Ph.D., Vice President for Animal Research Issues, The Humane Society of the
      United States
                                          IV

-------
                               TABLE OF CONTENTS
LIST OF FIGURES	vi
LIST OF TABLES	vi
ACRONYMS	vii
1. Introduction	1
2. Regulatory Applications and Impacts	5
  2.1 Chemical Screening and Prioritization	5
  2.2 Toxicity Pathway-Based Risk Assessment	5
  2.3 Institutional Transition	7
3. Toxicity Pathway Identification and Chemical Screening and Prioritization	8
  3.1 Strategic Goal 1: Toxicity Pathway Identification and Assay Development	10
  3.2 Strategic Goal 2: Chemical Prioritization	11
4. Toxicity Pathway-Based Risk Assessment	12
  4.1 Strategic Goal 3: Toxicity Pathway Knowledgebases	13
  4.2 Strategic Goal 4: Virtual Tissues, Organs, and Systems: Linking Exposure, Dosimetry, and
  Response	14
  4.3 Strategic Goal 5: Human Evaluation and Quantitative Risk Assessment	16
5. Institutional Transition	18
  5.1 Strategic Goal 6: Operational Transition	18
  5.2 Strategic Goal 7: Organizational Transition	20
  5.3 Strategic Goal 8: Outreach	20
6. Future Steps	23
Appendix: Other Related Activities	24
References	27
                                          v

-------
                                 LIST OF FIGURES

Figure 1. Toxicity Pathways	2
Figure 2. Toxicity Pathways Target Multiple Levels of Biological Organization	8
Figures. ToxCast™	1
Figure 4. Toxicity Pathways to Dose-Response	12
Figure 5. Knowledgebase Development	14
Figure 6. Relative (%) Emphasis of the Three Main Components of this Strategic Plan over its
        Expected 20-year Duration	23
                                 LIST OF TABLES

Table 1. Strategic Plan: Applications and Impacts	6
                                         VI

-------
                                    ACRONYMS

ACToR      Aggregated Computational Toxicology Resource
FIFRA       Federal Insecticide, Fungicide, and Rodenticide Act
FTTW       Future of Toxicity Testing Workgroup
HTS         High Throughput Screening
IRIS         Integrated Risk Information System
NRC         National Research Council of the National Academies
OPPTS       Office of Prevention, Pesticides, and Toxic Substances
ORD         Office of Research and Development
QSAR       Quantitative Structure-Activity Relationship
SAR         Structure-Activity Relationships
                                         vn

-------
                                     1.  INTRODUCTION

EPA bases its regulatory decisions on a wide range of tools and information that represent the best
available science. In some situations, where very limited or no animal toxicity data exist, EPA may
use tools such as structure-activity relationships (SAR) and quantitative structure-activity
relationship (QSAR) modeling, together with information on exposure to make decisions about
priority setting and the need for further evaluation (e.g., for new chemicals in the toxics program,
high production volume chemicals, and pesticide inerts). To establish regulatory standards, EPA
relies heavily on toxicity testing to evaluate clinical or pathological effects in experimental animal
models. As such, toxicity testing and related research is currently a multi-billion dollar activity that
engages thousands of research scientists, risk assessors, and risk managers throughout the world.
To that end, the historical path taken in toxicity testing of environmental agents has generally been
either to make incremental modifications to existing tests or to add additional tests to cover
endpoints not previously considered (e.g., developmental neurotoxicity). This approach has led
over time to a continual increase in the number of tests, cost of testing, use of laboratory animals,
and time to develop and review the resulting data. Moreover, the application of current toxicity
testing and risk assessment approaches to meet existing, and evolving, regulatory needs has
encountered challenges in obtaining data on the tens of thousands of chemicals to which people are
potentially exposed and in accommodating increasingly complex issues (e.g., lifestage
susceptibility, mixtures, varying exposure scenarios, cumulative risk, understanding mechanisms
of toxicity and their implications in assessing dose-response, and characterization of uncertainty)1.

While the challenges of such information gaps are great, the explosion of new scientific tools in
computational, informational, and molecular sciences offers great promise to address these
challenges and greatly strengthen toxicity testing and risk assessment approaches. Proven benefits
have been demonstrated in allied fields such as medicine and pharmaceuticals. Although untapped,
the potential  application to toxicity testing and risk assessment has also been recognized by EPA as
witnessed by the issuance of a series of papers that provided guidance on the use of genomic data.2
To better anticipate the potential contribution of new technologies and scientific advances to issues
associated with toxicity testing and risk assessment, EPA commissioned the National Research
Council (NRC) in 2004 to review existing strategies (NRC, 2006) and develop a long range vision
for toxicity testing and risk assessment (NRC, 2007). In the subsequent release of Toxicity  Testing
in the 21st Century: a Vision and a Strategy, a landmark transformation in toxicity testing and risk
assessment is envisioned that focuses on "toxicity pathways."3 This approach is based on the
rapidly evolving scientific understanding of how genes, proteins, and small molecules interact to
form molecular pathways that maintain cell function. The goal is to determine how exposure to
environmental  agents can perturb these pathways causing a cascade of subsequent key events
1 These limitations have been described more fully in .4 Review of the Reference Dose and Reference Concentration
Processes: http://www.epa.gov/ncea/iris/RFD_FINAL[l].pdf
2 Interim Policy on Genomics (2002): http://www.epa.gov/osa/spc/genomics.htm: Genomics White Paper (2004):
http://www.epa.gov/osa/pdfs/EPA-Genomics-White-Paper.pdf: Interim Guidance for Microarray-Based Assays
(2007): http://www.epa.gov/osa/spc/pdfs/epa interim guidance for microarrav-based assavs-external-
review draftpdf.
3 Toxicity pathways are cellular response pathways that, when sufficiently perturbed, are expected to result in
adverse health effects.

-------
leading to adverse health effects. This sequence of events is illustrated in Figure 1 wherein the
introduction of an environmental stressor may trigger such a cascade. Successful application of
these new scientific tools and approaches will inform and produce more credible decision making
with an increased efficiency in design and costs and a reduction in animal usage.
   Biologic
   Inputs
                   Source
                    *
                Fate/Transport
                    *
                  Exposure

                 Tissue Dose
                    *
               Biologic Interaction

                 Perturbation
   Toxicity Pathways: Cettuiar
     response pathways that,
     when sufficiently
     perturbed, are expected
     to result in adverse health
     effects.
Adaptive Stress
  Responses
                                           Mortality
 Modfied from NRC, 2007
Figure 1. Toxicity Pathways. Toxicity pathways describe the
processes by which perturbations of normal biological processes
due to exposure to a stressor (e.g., chemical) produce changes
sufficient to lead to cell injury and subsequent events (modified
from NRC, 2007).
Other agencies have also recognized
the need for this transformative shift,
including the National Toxicology
Program in their Roadmap for the
Future and the Food and Drug
Administration in their Critical Path
Program. In anticipating the
emergence, and potential, of this new
scientific paradigm, EPA's Office of
Research and Development (ORD) and
some of the Agency's regulatory
programs have also begun to redirect
resources in intramural and extramural
research programs to "jump start" the
process of transformation. For
example, ORD created the National
Center for Computational Toxicology4
in 2006. Likewise, ORD National
Laboratories and Centers  have also
begun to incorporate these new scientific tools to better support the research being conducted
under several of its multiyear research plans. Several ongoing projects address the use of in vitro
assays in risk assessment and toxicity testing (e.g., Guyton, etal., 2008), and assessments under
the Integrated Risk Information System (IRIS)5 program are describing and evaluating published
genomic data. EPA's Office of Prevention, Pesticides, and Toxic Substances (OPPTS) is also
actively involved in the development and transition of computational toxicology tools into
regulatory practice. OPPTS has developed a multi-year strategic plan to advance computational
toxicology tools in its risk assessment and management paradigm. Current activities include
assisting ORD by providing the necessary databases to support the development of models for
efficiently and credibly predicting toxic potency and levels of exposure, beta testing the new
computer models, training staff, and initiating plans for successful international coordination and
stakeholder involvement. Furthermore, recognizing the need to partner to achieve the vision and
goals laid out by the NRC, EPA recently signed a Memorandum of Understanding for research
cooperation with the National Toxicology Program and the National Institutes of Health
Chemical Genomics Center as a substantive step forward in building collaborations across sister
federal agencies.6 EPA is also working actively at the international level with programs such as
the Organization for Economic Cooperation and Development (OECD) through the Molecular
4 Computational toxicology is the application of mathematical and computer models and molecular biological
approaches to improve the Agency's prioritization of data requirements and risk assessments (from.4 Framework
for a Computational Toxicology Research Program, EPA 600/R-03/065).
5 http://cfpub.epa.gov/ncea/iris/index.cfm
6 http://www.epa.gov/comptox/articles/comptox mou.html

-------
Screening Initiative, the Integrated Approaches for Testing and Assessment Workgroup, Test
Guideline Committees, and the QSAR Expert Group to ensure global harmonization of any new
approach that originates from the research program. A more complete listing of these
collaborations may be found in the appendix.

In response to the release of the NRC reports, EPA has established an intragency workgroup, the
Future of Toxicity Testing Workgroup (FTTW), under the auspices of the Science Policy
Council. The FTTW includes representatives from across the Agency, including the Regions and
all major Program Offices. It has produced this current document, which will serve as a blueprint
for ensuring a leadership role for EPA in pursuing the  directions and recommendations presented
in the 2007 NRC report. This document presents a strategy that is consistent with the NRC's
directions and recommendations. It presents the Agency's vision of how to incorporate a new
scientific paradigm and new tools into toxicity testing  and risk assessment practices with ever-
decreasing reliance on traditional apical approaches. The overall goal of this strategy is to
provide the tools and approaches to move from a near  exclusive use of animal tests for predicting
human health effects to a process that relies more heavily on in vitro assays, especially those
using human cell lines. The topics to be covered include (1) the applications and impacts/benefits
for various types of regulatory activities (Section 2), (2) the research to be conducted to facilitate
the screening and prioritization of environmental agents (Section 3), (3) the implementation of a
toxicity pathway-based approach to risk assessment (Section 4), and (4) the critical companion
component, namely, the institutional transition that must occur before the changes can be fully
implemented (Section 5).

As described in Section 6, the workgroup recognizes that the full implementation of the vision
set out in this strategy will require a significant investment of resources over a long period of
time. The workgroup has identified a range of partners in this effort, and some planning on the
relative role of these partners has begun, although the specific areas of work to be
conducted/funded by EPA versus other partners needs further assessment. Decisions on the
relative roles will have a significant impact on EPA resources required to implement the vision.

Since the NRC charge and report centered on advancing toxicity testing for assessing human
health effects of environmental agents, this strategic plan is presented primarily within that
context. However, under environmental legislative mandates (e.g., the Toxic Substances Control
Act; the Federal Insecticide, Fungicide, and Rodenticide Act; and the Clean Water Act), most
EPA programs must regulate compounds to ensure both environmental and human health risks
are properly managed. Since statutory language and/or resulting policy typically require single
regulatory decisions for a chemical(s) that encompass  environmental and human health  risks at
the same time, accelerated and cost effective approaches for both areas are critical to realize
programmatic benefits. As in the human health arena, development and application of
approaches described in this strategy apply to ecotoxicology and risk assessment as well. Notable
progress is being made within EPA Laboratories and Centers on the development and use of
toxicity pathway models and the creation of prioritization schemes, toxicology knowledgebases,
and systems biology models in the field of environmental science.  The bringing together of
relevant disciplines to share data and integrate models is critical to fully achieve increased
efficiency in toxicity testing and a reduction in animal usage for both human health and
environmental risk assessment. Consequently, the Agency will be implementing this strategy in a
manner that addresses both human health and ecological risk assessment. Future versions of the

-------
strategy will summarize progress made in advancing integrated testing and assessment capability
and revisit remaining challenges.

-------
                   2. REGULATORY APPLICATIONS AND IMPACTS

The research arising from implementation of this strategy will change the nature of the methods,
models, and data that will inform the major components of the risk assessment process (i.e.,
hazard identification, dose response, exposure assessment, and risk characterization). Without
attempting to be all-inclusive, Table 1 presents some of the major cross-office applications and
impacts of these new scientific approaches, with more in-depth discussion of the planned work
described in Sections 3-5. The three components of this strategic plan, namely, chemical
screening and prioritization, toxicity pathway-based risk assessment, and institutional transition,
are not independent elements but rather highly interactive and integrative efforts that will
maximize the value and application of the research generated.

2.1.    Chemical Screening and Prioritization

An ongoing need of several regulatory offices is to have tools to assist in chemical screening and
prioritization, e.g., high production volume chemicals, air toxics, the drinking water Contaminant
Candidate Lists, and Superfund chemicals. These programs consider anticipated exposure and
hazard to select chemicals to evaluate in longer-term, whole-animal laboratory studies. An early
use for data developed under the new paradigm will be as an efficient and cost effective screen
for several types of chemical toxicity. Thus, risk assessors could use in silico (computer-based)
technologies and structure/molecular/bioactivity profiling from diagnostic high-throughput/m
vitro assays, along with predicted exposure/dose information, to predict chemicals most likely to
cause hazards of concern for humans. This approach will also enable risk assessors to determine
the specific effects, in vivo data, and exposures that would be most useful to assess, quantify, and
manage. As the technology develops, EPA will be able to screen previously untested chemicals
using libraries of chemical, molecular, biological, and toxicological data and models to identify
the types of adverse effects that they are most likely to produce in standard animal bioassays.
More importantly,  EPA will be able to gain better insight into whether such effects would likely
be manifest in humans under various exposure scenarios. As noted earlier, these needs are
common to a number of federal agencies; discussions are underway to develop more common
paradigms among federal agencies to facilitate data sharing.

2.2.    Toxicity Pathway-Based Risk Assessment

The current approach to risk assessment includes uncertainties associated with (1) the human
relevance of laboratory animal studies (species extrapolation), (2) the use of high doses in
animals to estimate risk associated with lower environmental/ambient exposures (dose
extrapolation), and (3) predicting the risk to susceptible populations. In recent years, the
consideration of such issues has been better informed by the incorporation of information on
potential modes of action through which toxicity may be expressed. The approach outlined
earlier in Figure 1 focuses on perturbations in baseline biological processes that may lead down
toxicity pathways to adverse health outcome(s). Combining this information with distributional
data on population characteristics of exposure and dose (magnitude,  frequency, and duration)
provides a scientifically based approach for reducing the uncertainties associated with current
risk assessments. By relying on a quantitative understanding of perturbations in toxicity
pathways that lead to adverse  health effects, the new approach to toxicity testing and risk
assessment envisioned in this  document will greatly increase EPA's  capacity to assess individual

-------
chemicals and their mixtures. The new approach will also increase EPA's confidence that the
Agency's assessments adequately protect human health. Realization and acceptance of this new
approach will likely encounter numerous challenges, but the effort is expected to ultimately lead
to better protection of human health.
                               Table 1. Strategic Plan: Applications and Impacts
         Toxicity Pathway
         Identification and
       Chemical Screening &
           Priori tization
                           Toxicity Pathway-Based Risk Assessment
                                                   Institutional Transition
      Need to screen 10,000's
      of chemicals for wide
      range of endpoints in a
      manner that considers
      toxicity pathways and the
      potential for human
      exposure.
                          For many chemicals, the current approach
                          relies on expensive animal testing that takes
                          time to conduct and review. Limitations in the
                          design of in vivo studies often prevent
                          complete evaluation of all endpoints and
                          hazard/risk scenarios of concern.

                          Limited understanding of biological
                          mechanisms most often leads to uncertainty in
                          assessing cumulative risk or extrapolating in
                          vitro to in vivo or across doses, lifestages,
                          species, or genetic diversity.
                                           Implementing the new approach will
                                           require significant institutional investment
                                           in operational and organizational transition
                                           and in public outreach.
 41
Need to limit cost and
animal usage, improve
timeliness, and decrease
uncertainty in testing
decisions.
New scientific understanding and tools in
molecular, computational, and information
sciences consistent with applications in allied
areas such as medicine and Pharmaceuticals
represent a path forward.
EPA lacks appropriate expertise and
sufficient funding to fully and most
efficiently utilize the new toxicity testing
technologies when making regulatory
decisions.
 o.
 I
 Z
Identification of toxicity
pathways for key
toxicological endpoints.
Combine in silica and
bioprofiles from HTS
along with QSAR
approaches linked to
animal study data.
Reliance on increased understanding of how
perturbations of biological processes at
environmentally relevant concentrations
trigger events (i.e., toxicity pathway(s)) that
may lead to adverse health outcomes.

Develop linked exposure/dose models to
inform dosing levels for toxicity testing and
inform risks.
Fully adopting the new paradigm should
be supported by mechanistically based
proof-of-concept and verification studies.
Further, such adoption will require
additional training of existing staff and
hiring new staff conversant in state-of-the-
science knowledge in fields such as
toxicology, biochemistry, bioinformatics,
etc.
 03
 O.
Offices would be better
able to direct efforts and
resources to chemicals
with greatest potential
risk. Significant increase
in efficiency with marked
reduction in cost for
toxicity testing.
                                More scientifically relevant data on which to
                                base EPA's regulatory decisions and/or
                                impact analyses that rely on these risk
                                assessments.
                                           A well informed public will have greater
                                           confidence as EPA greatly expands the
                                           number of chemicals assessed for possible
                                           risks and improves existing strategies for
                                           hazard and risk assessment!
7 High-Throughput Screening (HTS) refers to robotic technologies developed by the pharmaceutical industry for
drug development that enable the ability to evaluate the effects of hundreds to thousands of chemicals per day on
molecular, biochemical or cellular processes.

-------
2.3.    Institutional Transition

Implementing major changes in toxicity testing of environmental contaminants and incorporating
new types of toxicity data into risk assessment will require significant institutional change
involving:

   •   Operational transition - how EPA will transition to the use of new types of data and
       models for toxicity testing and risk assessment;
   •   Organizational transition - how EPA will deploy resources necessary to implement the
       new toxicity testing paradigm such as hiring of scientists with particular scientific
       expertise and training of existing scientific staff and risk managers;
   •   Outreach - efforts by EPA to share information with the public and improve risk
       communication.

The process of moving from research to regulatory acceptance for implementing new science
related to toxicity testing will be an iterative and long-term effort (likely encompassing more
than a decade). Essential to this iterative process will be the demonstration that the predictive
nature of these new approaches is superior to that of our current practices for toxicity testing and
risk assessment. It will be critical to begin activities geared toward regulatory acceptance early in
the process of implementing this strategic plan.

-------
      3.  TOXICITY PATHWAY IDENTIFICATION AND CHEMICAL SCREENING
                                    AND PRIORITIZATION

The advancements in biotechnology brought about by the sequencing of the human genome and
the investment in high throughput screening tools to mine large chemical libraries for potential
drugs have for the first time allowed a broad scale, unbiased examination of the molecular and
cellular targets of chemicals. At this time, the examination of the relationships between the
molecular and cellular targets of chemicals and the traditional endpoints of toxicity is at an early
stage of development. Even upon characterization of these types of relationships, significant
phenotypic data will be required to critically establish the role of toxicity pathways in evaluating
hazards and risks. The great potential is that identification of a toxicity pathway and
development of an in vitro bioassay for studying its chemical interactions will enable evaluation
of the effects of thousands of chemicals in that pathway. Broadening this approach to the many
toxicity pathways present in living systems allows a new avenue for identifying those chemicals
that pose the greatest potential hazard. Knowledge of the toxicity pathways triggered by any one
chemical will also allow targeting of specific in vivo tests to more fully characterize the potential
hazard and risk. The identification of toxicity pathways for key target tissues, organs, and
lifestages, and their linkage across levels of biological organization and exposure pathways  and
intensities are core elements of this  strategy.

As indicated in Figure 2, chemicals may interact with  a single pathway (the blue chemical) or
multiple pathways (the yellow chemical). Also, multiple pathways can lead to the same
expression of toxicity in the target organ as signaling pathways converge on common elements.
It is important to note that multiple
       r                             1—I     Chemicals , ,   _

                                     ********
Receptors / Enzymes / etc.
Direct Molecular Interaction
mechanisms of action  for any
particular adverse response likely
exist, and that many environmental
pollutants are likely to have multiple
mechanisms of action. Two critical
components of the toxicity pathway
concept are (1) extending knowledge
of molecular perturbations and cell
signaling pathways to understand
linkages between levels of biological
organization and (2) extending
knowledge of in vitro and in vivo
markers relevant to adaptive changes and/or adverse outcomes (see Section 5). As the research
moves forward, it will be important to capture quantitative relationships between the molecular
events and the higher order changes. Demonstration of plausible connectivity along the
mechanism of action from initiating event to adverse outcome will serve as the rationale for
using data from subcellular or cell-based in vitro assays for not only chemical prioritization but
also predictive risk assessment. As toxicity pathways are identified, relevant in vitro assays can
                                                                    Pathway Regulation /
                                                                    Genomics
                                                                Cellular Processes
                                                          Tissue / Organ / Organism Tox Endpoint
                                     Figure 2. Toxicity Pathways Target Multiple Levels of
                                     Biological Organization.
 Mode of action is defined as a sequence of key events and processes, starting with interaction of an agent with a
cell, proceeding through operational and anatomical changes, and resulting in an adverse health effect. Mechanism
of action implies a more detailed understanding and description of events, often at the molecular level, than is meant
by mode of action.

-------
be utilized and their results compared to in vivo studies as appropriate given the need to predict
effects in humans or other species. While comparing responses to those in animal bioassays will
be an early milestone of this strategy, the ultimate goal is the prediction of human risk.
Therefore, efforts will shift towards that goal as experience with the approach increases. An
added benefit to the toxicity pathway approach is that mixtures or their components could be
evaluated in this manner,  and as knowledge grows, it will be possible to predict where
interaction with multiple toxicity pathways might be expected to lead to non-additive outcomes.
This later activity will be  an important outcome of the research highlighted in Section 4.2
(Strategic Goal 4) that is focused on the development of virtual tissue models. As noted below,
virtual tissue models will  also provide a basis for predicting emergent properties of tissues by
integrating knowledge of molecular and cellular behaviors obtained from reductionist in vitro
approaches.

In 2007, EPA launched ToxCast™ 9 in order to develop a cost-effective approach for prioritizing
the toxicity testing of large numbers  of chemicals in a  short period of time. Using data from a
broad range  of state-of-the-art HTS bioassays developed in the pharmaceutical industry,
ToxCast   is building computational  models to forecast the potential human toxicity of
chemicals. Results from the HTS bioassays are being analyzed for signatures of bioactivity that
correlate with known toxicities.  These hazard predictions will provide EPA regulatory programs
with science-based information helpful in prioritizing chemicals for more detailed toxicological
evaluations,  and lead to more efficient use of animal testing.

The research described here focuses on two major strategic goals:
   1)  Identification of toxicity pathways and deployment of in vitro assays to characterize the
       ability of chemicals to perturb those pathways in different biological contexts, and
   2)  Implementation of ToxCast™, with an initial focus  on providing input for chemical
       prioritization, shifting over time to providing input for dose-response modeling.
A key feature of ToxCast™ is the phased nature of implementation (see Strategic Goal 2, Section
3.2), from proof of concept, to forward validation, and finally to reduction to practice. The
number of chemicals will grow from the hundreds to the thousands, and the number of assays
will change as experience and biology dictate. As the number of chemicals and breadth of
toxicity pathways covered increase, ToxCast™ will improve as a unique resource to build chemo-
informatic-based predictions of chemicals' potential human toxicity. Such advancements should
help  promote improved QSAR models and data upon which to build virtual tissue models.

Exposure science also plays a large role in this strategy. More simple and reliable screening
models are needed that predict exposures to chemicals so that information from the full source-
to-outcome continuum is  brought into consideration in the evaluation of chemicals  - a critically
important step for new chemicals that have not yet been released into the environment. Examples
of such simple methods and models for new chemicals can be found at EPA's Sustainable
Futures Initiative10. Additional such models  should further evaluate exposure based on the life
cycle of intended product use and the physical-chemical properties  of the chemicals. This
9 http://www.epa.gov/ncct/toxcast/
10 http://www.epa.gov/oppt/sf/

-------
research should include the expansion of computational chemistry methods to further predict
exposures as well as methods to predict release into the environment during product life cycle.
Several additional screening-level models are currently under development in Canada and
Europe. Research in this area should be coordinated with these groups to facilitate an
international approach for chemical screening. EPA should promote easy public access to all of
these additional models through the Internet.

3.1.    Strategic Goal 1: Toxicity Pathway Identification and Assay Development

The most systematic and extensive approach currently underway for screening and prioritization
is EPA's ToxCast™. Fully implementing the proposed strategy for more efficient toxicity testing
will utilize a combination of the more exploratory ToxCast  chemical signature approach (see
Strategic Goal 2), and the more hypothesis-driven approaches to elucidating toxicity pathways.
Developing  systems-based models will require comprehensive identification of the biological
processes that can result in toxicity when they are perturbed by chemical exposures. Therefore,
toxicity pathway identification and development of appropriate in vitro assays to characterize the
dose-response and time course of perturbations to those pathways will be needed. Measurement
of chemical  form and concentration from in vitro assays will also be important in hypothesis-
driven research that seeks to establish linkages between perturbations of toxicity pathways and
adverse effects, as well as for establishing structure-activity relationships. These research goals
will utilize a range of methods (e.g., transcriptomic, proteomic, metabolomic, cellular, and
biochemical analyses) to identify toxicity pathways using in vivo and in vitro systems. The in
vitro assays  and toxicity pathways already included in the ToxCast™ project will be a part of this
research, but additional assays providing greater coverage of relevant toxicity pathways will
need to be developed. For example, developmental neurotoxicity key responses are known to
include cell proliferation, apoptosis, differentiation (into different cell types and creating
different functionality/architecture of a cell), neurite outgrowth, synaptogenesis, and myelination
(Coeke et al., 2007; Lien et al., 2007), but the underlying molecular pathways are not yet
completely identified. Through the informed use of newer "systems-based" approaches (Edwards
& Preston, 2008), the flow of molecular regulatory information underlying the control of these
cellular events can be characterized, classified, and modeled. To facilitate use in risk assessment,
these studies will be coupled with mechanism of action-based studies, including animal and
human components as described in Strategic Goal 4.

Current priorities for research include developing in vitro assays for the key targets of chemicals
in the environment for which limited  knowledge is available (e.g., developmental neurotoxicity,
immunotoxicity, reproductive toxicity) as well as for relatively well-characterized toxicity
pathways such as stress response signaling. Studies representative of the full range of human
variability will be necessary to characterize processes that may occur more readily in sensitive
populations  (e.g., asthmatics) or at certain lifestages (e.g., prenatal development). Additional
emphasis needs to be placed on toxicities demonstrated to occur in humans.  For example, clinical
trials or post-marketing surveillance for pharmaceuticals, as well as molecular and genetic
epidemiology studies, afford the opportunity to examine effects of chemicals already introduced
into the environment that may not currently be well assessed by in vivo animal toxicity studies.
Some of these pathways may be important for environmental chemicals with respect to human
variability or exposure to complex mixtures.
                                            10

-------
-omics      Machine Learning
3.2.    Strategic Goal 2: Chemical Prioritization

This strategy extends approaches that are currently under development for EPA's ToxCast
program to include greater coverage of toxicity pathways and chemicals. The goal of the
ToxCast™ program is to provide a comprehensive assessment of toxicity pathways for a
relatively low cost per chemical (current estimates are in range of $20-25,000). ToxCast™ (see
Figure 3) was designed to collect
                                                   in vitro testing in silica analysis
data from a wide range or in vitro
assays, mostly mechanistic in nature,
to prioritize which chemicals to test          d  _
further and which in vivo studies
were likely most important. This
screening and prioritization approach                   HTS
        °     r             r r                       -omics
provides a near-term benefit during
an extended transition to the more     Figure 3. ToxCast™ is using a variety of HTS assays to develop
comprehensive proposed vision. As   bioactivity signatures that are predictive of effects in traditional
more comprehensive descriptions of   toxicity testing approaches.
processes involved in toxicological
responses become available, different assays may be identified to replace those in the initial
ToxCast™ effort, and the relationship to in vivo studies will shift from prioritization to providing
input for dose-response modeling.

ToxCast  is being developed in a phased manner. During FY08-09, substantial progress will be
made on the first two phases of the ToxCast™ program (Dix et al., 2007; Kavlock et al., 2008).
Phase I is a proof of concept involving 320 chemicals that have robust in vivo animal toxicity
information. These chemicals have been profiled using over 400 high and medium throughput in
vitro assays. From these in vitro bioactivity profiles, classifiers or signatures predictive of
chemicals'  in vivo toxicity are being derived. Phase II will  involve validation of the predictive
bioactivity  and expansion of the diversity of chemicals tested. Phase III is the most relevant to
this strategic plan, as it would begin to apply the knowledge gained in Phases I and II to the tens
of thousands of chemicals of concern to EPA regulatory offices. An adaptation of the approach
to evaluate the hazardous properties of nanomaterials is also anticipated.
                                            11

-------
                  4.  TOXICITY PATHWAY-BASED RISK ASSESSMENT

The goals of the proposed new strategy for toxicity testing include collecting mechanistic data,
largely in vitro, for the purpose of predicting human risk from exposure to chemicals. Prediction
of in vivo effects in humans requires a combination of measurements and computer modeling to
link in vitro responses to tissue dosimetry to alterations in the structure and function of tissues
and organs. A substantial challenge will be to address the range of human variability arising from
differences in age, life stage, genetics, disease susceptibility, epigenetics, diet, disease status, and
other factors that potentially influence or interact with toxicity pathways.

The initial process for predicting human risk under this new approach could be summarized as
(1)  characterizing or predicting potential human exposures; (2) estimating the resulting chemical
dosimetry (magnitude, frequency, and duration) for target pathways, tissues or organs; (3)
measuring toxicity pathway response at doses consistent with human exposures; (4) predicting
the  in vivo human response resulting from pathway perturbations; (5) quantifying the range of
human variability and susceptibility; and (6) validating predictions utilizing in vivo systems (e.g.,
laboratory animals, human data). In the current state of mechanistic toxicology (top row of
Figure 4), chemicals are administered to the test animals (usually at high doses), a variety of
        Envi nan mental
         Chemicals
Mo le cu la r
 Sensing
 Cellular
Signaling
  Tissue
Responses
          *.*


      Knowledgebase
         Toxicity
         Path ways
Molecular
Networks
 Cellular
Networks
 Virtual
Tissues
Dose-Response
  Figure 4. Toxicity Pathways to Dose-Response. The vertical arrows at each step in the process reflect the
  iterative nature of experimentation and modeling needed to gain full understanding of both the toxicity pathway
  determination and the relationship to normal biology.

biochemical approaches are used to detect alterations in molecular pathways, the data are mined
to describe the ensuing cellular alterations (e.g., oxidative stress damage, mitochondrial
dysfunction), and tissue changes are confirmed at the level of morphology or function. The
                                             12

-------
bottom row of the figure depicts the vision for future ways of assessing risk, which includes
determining the key toxicity pathways, defining approaches for examining perturbations in
molecular networks, and translating the results to responses at the cell, and ultimately tissue and
organ level, using computational models of the relevant systems. The expectation is that
assessments in the future will utilize data from in vitro studies, and the need for in vivo animal
testing will be substantially reduced. However, until the state of science of this new approach has
reached a level of confidence for use in regulatory decision making, the traditional  approach to
toxicity testing will continue into the foreseeable future. With time, we expect that  it will be
progressively augmented and ideally replaced by computational models that integrate the
information generated from non-animal sources into predictive models of response based upon
the underlying biology. The vertical arrows at each step in the process reflect the iterative nature
of experimentation and modeling needed to gain full understanding of both the toxicity pathway
determination and the relationship to unperturbed biology. One anticipated outcome of the
development  of virtual tissues will be an increased understanding of the role of metabolism and
of intra- and inter-cellular signaling pathways. This understanding will lead to the development
of improved in vitro systems that, for  example, might include combined cell-based  systems to
provide metabolic competency or to better reflect the intercellular responses in heterogeneous
tissues.

As the transition progresses, it is important that increased emphasis will be placed on
examination of exposure concentrations that are expected to occur in the environment. The key
difference in future toxicity evaluations will be the transition to a focus on ways in  which
molecular pathways (as detected by in vitro models) are perturbed by chemical exposure
throughout the range of exposures  from environmental to the higher dose levels commonly used
in contemporary toxicity studies. Dosimetry measurements coupled with computational
modeling will be critical for predicting in vivo exposure levels of concern and for determining
relevant in vitro concentrations. Some responses of targeted toxicity pathways can be evaluated
in simpler cell culture models, whereas, in other cases, multiple in vitro assays may be necessary
for the integration of multiple pathways that produce in vivo responses. These situations would
require biologically based models for  the responses as well as for chemical dosimetry in  order to
predict the integrated in vivo response.

Implementing this new paradigm requires organization of existing scientific information;
computational methods for exposure,  chemical dosimetry, and perturbations of biological
processes; and evaluation of the methods for risk assessment applications. The research program
to implement this element of the strategy is defined by three goals: development of toxicity
pathway and exposure knowledgebases; development of virtual tissues, organs, and systems; and
evaluation of human relevance.

4.1.    Strategic Goal 3: Toxicity Pathway Knowledgebases

The underlying basis of the 2007 NRC report is that there are a finite number of toxicity
pathways (i.e., in the hundreds) that could be queried using in vitro assays to obtain insights into
the ability of chemicals to perturb those pathways. It refers to several  stress pathways (e.g.,
oxidative stress response) and notes the general listing of signaling pathways in a previous NRC
report (2006). However, an inventory of toxicity pathways and their involvement in a variety of
toxicological  responses needs to be created. Likewise, from exposure science there needs to be a
                                           13

-------
           ACToR API
                      Chemical ID.
                      Struclure
                       MicroArray
                       Data
Specialized Toxicology Databases
Biological
Reference
Data
Data Mining
complementary effort focusing on those chemical properties and computational methods that
could be used to reliably predict behaviors in the environment and exposures. This effort would
include information on stability in the environment, likely routes for exposure, potential for
bioaccumulation, and extent of metabolism. Therefore, a strategic goal is the development of a
knowledgebase for toxicity pathways and exposure. Knowledgebases differ from traditional
databases in the extent of integration of information and the inclusion of tools that can draw
inferences from amongst the diverse elements.

The knowledgebase would serve a variety of functions throughout the research and development
effort associated with implementing this new approach to toxicity testing and will become a
standard tool in the risk assessments of the future. ACToR (Figure 5), the Aggregated
                                                                 Computational
                                                                 Toxicology Resource
                                                                 under development in
                                                                 ORD, is an example of
                                                                 the needed approach of
                                                                 bringing together diverse
                                                                 types of information into
                                                                 a system where
                                                                 interrelationships of
                                                                 individual database
                                                                 elements (e.g., traditional
Figure 5. Knowledgebase Development. ACToR brings together a diverse set of  toxicology, chemical
currently unlinked resources available from internal and external sources into a    structure information,
system with a user friendly interface to readily mine and analyze toxicity data.     high throughput
                                                                 screening data, molecular
pathway analysis, chemical data repositories, peer reviewed published literature, and internal
Agency databases) can be explored and utilized  (Judson et al., 2008). Key  steps in development
of these knowledgebases include: (1) creating electronic repositories of existing toxicity
information; (2) developing semantics for describing toxicity pathways; (3) automating pathway
inference tools to aid in discovering mechanistic links between genomic information and
molecular and cellular observations; and (4) creating a toolbox with a user-friendly interface to
organize, access, and analyze toxicity pathway assay results.

4.2.   Strategic Goal 4: Virtual Tissues, Organs, and Systems: Linking Exposure,
       Dosimetry, and Response

Computational techniques relevant to this strategy fall into two general branches: knowledge-
discovery  (data-collection, mining, and analysis) represented in Strategic Goal 3, and dynamic
computer simulation (mathematical modeling at various levels of detail) described in this
section. The central premise of the latter approach is that critical effects of environmental agents
on molecular-, cellular-, tissue-, and organ-level pathways can be captured by computational
models that focus on the flow of molecular regulatory information (Knudsen & Kavlock, 2008).
This information flow is influenced by genetic and environmental signals,  with the net outcome
being the emergent properties associated with baseline or abnormal collective cell behavior.
Thus, computational systems modeling will be used to predict organ injury due to chemical
exposure by simulating: (1) the dynamics and characteristics of exposure and dose, (2) the
                           14

-------
dynamics of perturbed molecular pathways, (3) their linkage with processes leading to alterations
of cell state, and (4) the integration of the molecular and cellular responses into a physiological
tissue model. By placing a strong emphasis on understanding the biology of the system and the
key regulatory components, these virtual tissue models represent a significant opportunity to
better understand the linkage between chemically induced alterations in toxicity pathways and
effects at the  organ level. This research represents an ambitious effort, conceivable for the first
time due to the current technological advances. Virtual tissue and organ system models will
initially include liver, cardiopulmonary function, selected immune system tissues, multi-organ
endocrine axes, and developing embryonic tissues. Development of these virtual tissue and  organ
systems will require newly generated data to both fill data gaps identified within the iterative
process and test the predictive nature of these virtual systems. Comparative studies should
include pathways fundamentally reliant upon cell signaling (e.g., cell proliferation, apoptosis,
cell adhesion), intermediary metabolism (e.g., glycolysis, oxygen utilization, fatty acid
biosynthesis), differentiation-specific functions (e.g., extracellular matrix remodeling), and  other
categories as  developed above (see Strategic Goal 1) to ensure that predictions are broadly
applicable.  The wealth of existing data from NTP assays, published reports, and previous EPA
intramural studies will be leveraged wherever possible with additional experiments designed to
fill data gaps. Such efforts will also help answer how well in vitro experimental systems
represent the  full range of diverse cells present in the human body, how variability observed in
the human population can modify quantitative predictions of in vivo dose-response, how
exposure conditions influence outcomes, and how well the virtual tissue models represent the
underlying processes.

Not all toxicity pathways are likely to be expressed in every tissue, and likewise not all tissues
are likely to manifest adverse outcomes following chemical perturbation. Chemicals that affect
the same toxicity pathway can do so via a number of different (and overlapping) mechanisms,
and development of assays across toxicity pathways leading to the same outcome is a necessary
component of the proposed strategy. Some toxi cities are manifest only when multiple  cell types
and specific cell-cell interactions are present. Other toxicities may be dependent upon tissue
geometry and three-dimensional architecture. Examples include signaling between hepatocytes
and Kupffer cells, or the many forms of signaling between epithelial and mesenchymal cells. As
such, developers of virtual cells, tissues, organs, and systems must always bear in mind the  need
to remain relevant to the processes critical to expressions of toxicity in vivo. Consistent with the
NRC vision (2007), this need will likely entail a continued although decreasing role for in vivo
systems for the foreseeable future.

A premise of the new toxicity testing strategy is that computational methods combined with an
understanding of biological and exposure processes can be used to develop a more efficient and
accurate approach for predicting risks from many chemicals. On the exposure side, models have
been developed and are available that predict fate and transport, environmental concentration,
exposures,  and doses. These models work at multiple scales; for multiple sources, routes, and
pathways; and for multiple chemicals, although each model only addresses a single process or
compartment. Research is needed so that such models can take into account weathering of
contaminants, differences in bioavailability of contaminants, variations in exposures with age,
and variability in exposures within populations. Research is also needed to combine these models
across various scales to develop a linked source-to-outcome modeling framework, to evaluate the
framework using multiple chemicals and exposure scenarios, and to improve the computational
                                           15

-------
efficiency for the approach. Ultimately, these exposure models will be linked to the virtual tissue
models for utilizing in vitro toxicity test results in quantitative risk assessments. Given the
complexity of the challenges present in addressing each of these components, this effort
represents a long-term goal of the strategy. However, efforts must begin now to put us on the
path to achieving the ultimate vision of Toxicity Testing in the 21st Century (NRC, 2007).

The derived computational models must accurately describe the processes and mechanisms that
determine exposure and effect. They must have reliable input parameters in order to quantify
these processes. On the exposure side, our current understanding of processes and factors for
many classes of chemicals and pathways (i.e., dermal and incidental ingestion) is limited. New
approaches will be evaluated that will allow us to address the most significant uncertainties.
Relational databases populated with data on exposures, exposure factors, activity patterns, and
biomarkers will be developed as described. Informatic approaches or applications of network
theory could potentially be used to provide a better understanding of important exposures, as
well as exposure/response relationships. In the 2007 NRC report, emphasis was placed on
biomarkers and their role in relating real world exposure to in vivo and in vitro biological
response. They were also proposed as primary indicators in surveillance programs for tracking
predicted exposures and health outcomes. Because of this emphasis, novel approaches for using
biomarkers and integrating them into new risk assessment approaches will be investigated for
chemicals already existing in the human environment. Perhaps such biomarker data can be used
to improve predictive exposure models that will be  relied upon for new chemicals not yet
introduced into the environment.

4.3.   Strategic Goal 5: Human Evaluation and Quantitative Risk Assessment

The critical challenge of this new vision for toxicity testing using mechanistic in vitro assays,
targeted in vitro or in vivo testing, and computational models is to demonstrate that it
successfully and adequately predicts human toxicological responses. Proof of concept efforts
need to address this challenge both retrospectively and prospectively. Existing human data from
pharmaceutical and environmental studies will be used to the extent possible. Human data could
come from a range of sources including case reports, epidemiological studies (e.g., from the
National Children's Study), and clinical trials. EPA has extensive experience obtaining human
clinical data following exposure to the criteria air pollutants (e.g., ozone, particulate matter) and
other chemicals (e.g., MTBE)11. Engagement of the pharmaceutical industry and the Food and
Drug Administration to access toxicity findings from clinical trials of drugs that were
successfully registered or that failed to be registered would be a desirable component of this
effort.  Limited data may be available for some nutrients or dietary supplements as well.

Such efforts will help address the question of the extent to which key events (critical
perturbations) that are predictive of health endpoints (e.g., cancer, immunosuppression, kidney
disease) must be demonstrated or whether the perturbation of baseline biological processes
sufficient to induce substantial cellular level response (e.g., a stress response) should be
considered an adequate endpoint for risk assessment. Linking a specific pathway perturbation to
11 All EPA conducted or supported research is subject to and must comply with EPA regulations on the protection of
human subjects. See http://www.epa.gov/fedrgstr/EPA-GENEPvAL/2006/Februaiy/Day-06/gl045.htm:
http://www.epa.gov/oamrtpnc/forms/1000  17a.pdf
                                            16

-------
a particular target organ endpoint has the advantage of predicting outcomes that are already used
in risk assessment, while alternative approaches raise issues of which endpoints should and
should not be considered for risk assessment. This approach is relatively straightforward for
some effects (e.g., hemolysis of red blood cells by EGBE, where the effect and the mechanism of
action leading to it are qualitatively the same, even if quantitatively different). Linkage is more
complicated for effects observed in animals that may predict human effects that are related, but
not identical to, the outcomes in animals (e.g., developmental effects in an animal model may
predict developmental effects in humans, but the exact manifestation might be different).  On the
other hand, as knowledge is gained about the interaction of chemicals with molecular targets, and
this knowledge is combined with information on how perturbations of those targets are translated
to responses in species-specific patterns (e.g., how activation of certain transcription factors lead
to species-specific tissues responses), it will be increasingly possible to predict human outcomes
from in vitro studies that identify mechanism of action. Clearly this aspect will need to be
addressed on a case-by-case basis as we gain experience.

To be most useful in evaluation of risk to humans, the pathway-based efforts should ideally be
tied to a known mechanism of action, such as via the use of quantitative biologically based, dose-
response models. Understanding of the relevant mechanism of action will enable the
identification of biomarkers for key event parameters (linked to toxicity pathways) that can be
monitored in human studies for those chemicals already released into the environment at
significant levels. These biomarkers could be measured in observational human studies to
provide in vivo data to support the underlying pathway-based model. In addition, genetic
susceptibility in humans identified via whole genome association studies will provide support for
pathway-based models when genes critical for a key toxicity pathway are  associated with
susceptibility. Finally, the use of quantitative models requires estimation of uncertainty and
variability in the predictions from in vitro assays and computational models. Formal methods for
model evaluation are essential for demonstrating the success  of this new approach to toxicity
testing and risk assessment.
                                           17

-------
                            5. INSTITUTIONAL TRANSITION

Implementing major changes in toxicity testing of environmental contaminants and incorporating
new types of toxicity data into risk assessment will require significant institutional changes. This
section will touch upon three major thrusts of implementing institutional transition: operational
transition, organizational transition, and outreach.

5.1.    Strategic Goal 6: Operational Transition

Operational transition covers the technical aspects associated with EPA's implementation of a
new toxicity testing paradigm and associated changes in risk assessment. It will consider such
disparate topics as the importance of grounding the science, ensuring consistency of approaches
within EPA, and working with outside partners and issues associated with the use of new models
and tools.

The NRC "envision[s] a future in which tests based on human cell systems can serve as better
models of human biologic responses than apical studies in different species." Achieving such a
future, however, will require substantial research to study and define various toxicity pathways.
In evaluating possible options for the future of toxicity testing, the NRC eventually chose an
option involving both in vitro and in vivo tests but based primarily upon human biology and the
attendant use of substantially fewer animal studies that would be focused on mechanism and
metabolism. Their vision for the next 10 to 20 years relies on understanding perturbations of
critical cellular responses and the use of computational approaches for assessing hazard and risk.

A paradigm shift in toxicity testing based on pathway perturbation will likely require significant
methodological advances and future changes to EPA's risk assessment guidelines. Although it is
infeasible to denote a specific timeline for how long it will  take to substantially complete the
strategic goals associated with toxicity pathway identification, chemical screening and
prioritization, and toxicity pathway-based risk assessment,  this plan takes the view that advances
are likely to be gradual over the next decade or two. The good news is that toxicity testing
research efforts have already begun moving EPA and others towards the use of in silico
technologies and high throughput testing systems. The speed at which we are able to complete
this transition will depend on the availability of increased research funding. It is important to
note that our understanding of toxicity pathways for some apical endpoints (e.g., hepatotoxicity)
may be developed at a faster pace than others (e.g., neurotoxicity) thus, allowing more rapid
introduction of newer high-throughput in vitro testing methods.

Grounding the Science - From a broad regulatory perspective, data used by EPA to support
regulatory decisions will be shaped by the statutory language covering the action, regulatory
policies, and the resulting time and resources allocated to the assessment. Where appropriate, use
of data should be consistent with the EPA guidance articulated in a number of science policy and
guidance documents, including toxicity testing guidelines, risk assessment guidelines12,
information quality guidelines13, and peer review guidance.14
12 http ://www. epa. gov/risk/guidance. htm
13 http://www.epa.gov/qualitv/informationguidelines/
14
  http ://www .epa. gov/peerreview/pdfs/Peer%20Review%20HandbookMay06 .pdf


                                            18

-------
To implement this new paradigm, regulators, stakeholders, and the public will need to develop
confidence that the data generated can be used effectively and that public health will continue to
be protected. A step-wise implementation is envisioned: first, experience will be gained from
proof of concept studies using data from chemicals (e.g., pesticides) with a large set of toxicity
data developed using the current paradigm. Availability of both new and traditional types of data
will allow extrapolation and comparison of results across methodologies.

Optimally, early  success stories that meet programmatic needs in specific areas such as
mechanism of action analyses or cumulative risk assessments will demonstrate the broader
applicability of computational toxicology within the Agency. Reliability of the testing paradigm
will need to be evaluated via a comprehensive  development and review process, involving public
comment, harmonization with other agencies and international organizations, and  peer review by
experts in the field. Bringing new methods into regulatory practice will require several phases
starting from the development of the science and technologies, to technology transfer and
building the regulatory infrastructure, to incorporation of the new tools into decision making.

Because this transformative paradigm will rely on new and complex science and will  likely be
surrounded by some controversy, an important part of regulatory acceptance will be to conduct
research that will verify the approaches and models that will come to replace much of the way
toxicity testing and risk assessments are conducted in the Agency today. An  important
component of the effort to develop new approaches to testing will be to translate the research
into regulatory applications.

Issues Associated With the Use of New Methods and Models - For this new paradigm to be
successful, new methods and models should be thoroughly evaluated prior to their application
and use in regulatory decision making. The computer-based models used by  the Agency should
be publicly available. Testing methods should  be accompanied by documentation  that describes
(1) the method and its theoretical basis, (2) the techniques used to verify that the method is
accurate, and (3) the process used to evaluate whether the method and the results are sufficient to
provide an adequate basis for its use in regulatory decision making. Access to data to allow for
third party independent replication of results, to the extent practicable, is essential. Such review
is appropriate before the Agency relies on data from such a method.15

Working With Outside Partners -The appendix provides details about the many outside parties
EPA will need to partner with in order to implement this strategic plan including:

   •   Other federal bodies such as the National Toxicology Program (NTP) and  the NIH
       Chemical Genomics Center  (NCGC), with whom EPA has a memorandum of
       understanding to collaborate;
   •   The Interagency Coordinating Committee on the Validation of Alternative Methods
       (ICCVAM), which is made up of representatives from 15 federal agencies  that generate
       or use toxicological data;
   •   Foreign governmental parties and programs such as REACH, which is the  new European
       Union Regulation on Registration, Evaluation, Authorization, and Restriction of
       Chemicals that went into effect June 1, 2007;
 1 See http://epa.gov/crem/library/CREMguidancedraftl2 03.pdf


                                           19

-------
   •   The OECD (Organization for Economic Co-Operation and Development), which
       represents over 30 countries in the Americas, Europe and Asia;
   •   Academia;
   •   Chemical industry; and
   •   Non-governmental organizations.

Case Study Development - Significant challenges, such as interpretation and communication of
data obtained using new toxicity testing approaches, will emerge under a new paradigm for
toxicity testing. A key feature of a successful communication strategy will be to develop case
studies using new kinds of data that can serve as a basis to explore, evaluate, and most
importantly explain hazard, dose-response, and exposure information in a risk assessment
framework. Characterization of risk information, both qualitative and quantitative, in a manner
suitable for communication to risk managers will be a significant challenge for the research and
risk assessment community, but it will be crucial if the new toxicity testing paradigm is to reach
its potential.

5.2.    Strategic Goal 7: Organizational Transition

Organizational transition is meant to cover changes in direction over time with regard to
deployment of human capital resources necessary to implement the new toxicity testing
paradigm such as  hiring of scientists with particular scientific expertise and training of existing
scientific staff. For example, EPA has hired key new scientific staff and initiated training
including three new training courses in genomics designed and implemented by EPA's Risk
Assessment Forum. Additional resources and training programs will be needed in both EPA's
research program  as well as its regulatory and regional programs.

As noted in Section 2, several intra-agency, interagency, and international activities are already
underway to begin the transformation that will change the nature of toxicity data generated and
how it is used to assess chemically induced risks to human health. Substantial funding will be
needed to provide the scientific basis for creating new testing tools; to verify the utility of new
testing tools including conducting peer review; to develop and standardize data-storage, data-
access, and data-management systems; to evaluate  predictive power for humans;  and to improve
the understanding of the implications  of test results and how they can be applied in risk
assessments used  in environmental decision-making.

EPA expects that the use of less expensive, high-throughput testing methods will allow for the
generation of toxicity data for thousands of currently untested or under-tested chemicals. The
availability of these new data will likely lead to the need for more staff to interpret the data for
many more chemicals and manage their risks. Additionally, toxicity databases such as EPA's
IRIS and models used to assess risks may need to undergo substantial changes in the long term
requiring future resources.

5.3.    Strategic Goal 8: Outreach

Outreach consists of those efforts that will be used  to help educate the public and stakeholders as
well as improve risk communication.
                                           20

-------
In reaching out to the public, it will be important to re-emphasize points made by EPA
Administrator Carol Browner in a 1995 memorandum to senior Agency staff about the Agency's
policy related to its new Risk Characterization Program. This memorandum described the
importance of adhering to the "core values of transparency, clarity, consistency, and
reasonableness (which) need to guide each of us in our day-to-day work; from the toxicologist
reviewing the individual (scientific) study, to the exposure and risk assessors, to the risk
manager, and through to the ultimate decision-maker." Further, "because transparency in
decision-making and clarity in communication will likely lead to more outside questioning of our
assumptions and science policies, we must be more vigilant about ensuring that our core
assumptions and science policies are consistent and comparable across programs, well grounded
in science, and that they fall within a 'zone of reasonableness.'"16

Stakeholder Involvement - Implementation of a paradigm shift in toxicity testing and related
changes to risk assessment methods and practices will require a  sustained effort over many years
- remember that the NRC envisioned some 10 to 20 years to reach their goal. This transition to
new methods and approaches will need to be transparent, including efforts to share information
with both the public and risk managers. It will be critical to effectively communicate with
stakeholders (the public, scientists, federal and state agencies, industry, the mass media,
nongovernmental organizations) about the new tools and the overall program regarding its
strengths,  limitations, and uncertainties. One way to enhance stakeholder involvement and ensure
cooperation is to hold periodic workshops where all parties can gather to share information and
progress; another tool is for EPA to establish a web portal to detail advancements in the science
and relate these to improvements in risk assessment methods and practice.

Collaboration among different elements in the research community involved in relevant research
on new testing approaches will be needed to take advantage of the new knowledge, technologies,
and analytical tools as they are developed, and collaboration between  research and regulatory
scientists will be vital to ensure that the methods developed can be reliably used in risk
assessments of various types (initially qualitative, but ultimately both  qualitative and
quantitative). Mechanisms for ensuring sustained communication and collaboration, such  as data
sharing, will also be needed.  Independent review and evaluation of the new toxicity testing
paradigm should be conducted to provide advice for midcourse corrections, weigh progress,
evaluate new and emerging methods, and make any necessary refinements in light of new
scientific challenges/advances. This may be  accomplished using existing EPA mechanisms for
peer review, e.g., through reviews by the Board of Scientific Counselors, the Science Advisory
Board, and the FIFRA Scientific Advisory Panel. For testing that the Agency may wish to
require, performance standards should be considered so that individual methods from any
qualified source may be used. The NRC (2007) stressed that "in vitro  tests would be developed
not to predict the results of current [animal] apical toxicity tests but rather as [human] cell-based
assays that are informative about mechanistic responses of human tissues to toxic chemicals.
The [NRC] committee is aware of the implementation challenges that the new toxicity-testing
paradigm would face." Presumably, establishing regulatory confidence that the new approaches
are robust and protective of human health will be at the forefront of future challenges for EPA
and its partners.
16 http://www.epa.gov/oswer/riskassessment/pdf/1995 0521 risk characterization_program.pdf


                                           21

-------
Risk Communication - Communicating with policy makers and the public is an important part
of any risk management exercise. The complexity of the emerging toxicity testing paradigm and
how new types of data and information will be used to assess risk will make communication of
results challenging; consequently, the Agency must work to build public trust in the adopted
technologies. As the science moves away from well-established animal models, a significant
effort must be made to share information with risk assessors/managers and the public by clearly
describing test results and methodologies in a transparent manner.  A fundamental aspect of
gaining public trust is transparency. Therefore, education and effective communication with
stakeholders (the public, scientists, regulatory authorities, industry, the mass media, and
nongovernmental organizations) on the strengths, limitations, and uncertainties of the new
tools/paradigm will be critical.

Given that these new methods will be less intuitive than looking for traditional effects in whole
animal studies, communication strategies will be very important. At this time, much of EPA's
effort in this area is presented on the Agency's National Center for Computational Toxicology
Web site.17 As the new toxicity testing paradigm continues to evolve, the Agency will need to be
vigilant in maintaining an interactive Web site to describe each individual assay or method in use
and where it fits into the exposure-response continuum.

When communicating about risk, it is important for the Agency to address the source,  cause,
variability, uncertainty, and the potential adversity of the risks, including the degree of
confidence in the risk assessment methodology, the rationale for the risk management decision,
and the options for reducing risk (U.S. EPA, 1995; U.S. EPA,  1998). EPA will continue to
interact with stakeholders in order to develop and maintain effective informational tools.
17 http://www.epa.gov/comptox/


                                           22

-------
                                    6.  FUTURE STEPS
This strategic plan describes an ambitious and substantive change in the process by which
chemicals are evaluated for their toxicity. The NRC (2007) suggested that such a transformation
would require up to $100M per year in funding over a 10-20 year period to have a reasonable
chance of reaching the goals. Even including the resources of sister agencies, the overall federal
budget for the collaborative efforts does not approach the NRC proposed level of funding.
Decision on the relative role of EPA vis-a-vis other partners will have a major impact on the
resources that EPA needs to dedicate to this effort.  These decisions will have to be made as the
strategy is implemented. Explanation of these decisions, their rationale, and implications will be
included in a subsequent implementation plan.

Regardless of whatever  level of funding is ultimately applied to the vision of a more efficient and
effective chemical safety evaluation effort, translation of this strategy into research and activities
related to operational and organizational change will require development of an implementation
plan as well as periodic  peer review of directions and progress. Representatives from those EPA
organizations most involved and impacted by the new vision will play key roles in the
implementation program. The Science Advisory Board and/or the Board of Scientific Counselors
will play key roles in the scientific peer review of the program. As noted in Section 4, there will
be a progression in the
implementation efforts from an
early focus on hazard
identification to a growing
emphasis on the use of toxicity
pathway characterization in risk
assessment. Support for
institutional transitions is also
expected to increase over time as
the tools and technologies
emerge out of the research
programs and become available
for regulatory use. Figure 6
depicts one potential way that the
 •e
 £
                   —Screening/P rioritization
                   ^Toxidty Pathways in
                     Risk Assessment
                     Institutional Transi f on
      2010
2015    2020
   Year
2025
level of effort of the three main
activities involved in this strategy
could change over time.
Figure 6. Relative (%) Emphasis of the Three Main Components of
this Strategic Plan over its Expected 20-year Duration.
                                            23

-------
                     APPENDIX: OTHER RELATED ACTIVITIES

Other US Government Activities

The National Toxicology Program (NTP) at the National Institute of Environmental Health
Sciences (NIEHS) coordinates toxicological testing programs within the Department of Health
and Human Services18. Similar to EPA, NTP is developing the use of computational models, in
vitro assays, and non-mammalian in vivo assays targeting key pathways, molecular events, or
processes linked to disease or injury for incorporation into a transformed chemical testing
paradigm.

The NIH Chemical Genomics Center (NCGC) of the National Human Genome Research
Institute conducts ultra high throughput screening assays as part of the NIH's Molecular
Libraries Initiative within the NIH Roadmap

A Memorandum of Understanding19 was recently signed by EPA, the NTP, and the NCGC to
collaborate on generating a comprehensive map of the biological pathways affected by
environmental chemical exposures and use this map to predict how potential chemical toxicants
will affect various types of cells, tissues, and individuals. The hope is to refine many of the
toxicity tests performed on animals and eventually supplant them with in vitro testing and
computational prediction (Collins et al., 2008).

In 2004 the Food and Drug Administration (FDA) produced a report20 addressing the need to
translate the rapid advances in basic biomedical sciences into new preventions, treatments and
cures. FDA holds large databases of human, animal, and in vitro data for screening drug
candidates for toxicity that may also be useful for screening environmental chemicals. The
FDA's National Center for Toxicological Research (NCTR) aims to develop methods for the
analysis and integration of genomic, transcriptomic, proteomic, and metabolomic data to
elucidate mechanisms of toxicity21. NCTR has coordinated the Microarray  Quality Control
(MAQC) project, with numerous partners including EPA (Shi et al., 2006). In addition, NCTR
has provided its Array Track database to EPA for storage of genomics data for research and
possible regulatory use.

The Interagency Coordinating Committee on the Validation of Alternative Methods
(ICCVAM) was established by law in 2000 to promote  development, validation, and regulatory
acceptance of alternative safety testing methods. ICCVAM is made up of representatives from 15
federal agencies that generate or use toxicological data. Emphasis is on alternative methods that
will reduce, refine,  and/or replace the use of animals in testing while maintaining and promoting
scientific quality and the protection of human health and the environment22. The NTP
Interagency Center for the Evaluation of Alternative Toxicological Methods (NICEATM)
administers and provides scientific support for ICCVAM. ICCVAM/NICEATM evaluates test
method submissions and nominations, prepares technical review documents, and organizes
18 http://ntp.mehs.mh.gov/ntp/mainjages/NTPVision.pdf
19 http://www.epa.gov/ncct/articles/comptox mou.html:
20 http://69.20.19.211/oc/initiatives/criticalpath/whitepaper.html
21 http://www.fda.gov/nctr/overview/mission.htm
22 http://iccvam.niehs.nih.gov/about/ni QA.htm
                                          24

-------
scientific workshops and peer review meetings. For example, ICCVAM/NICEATM recently
               .23
released a report  that describes two in vitro cytotoxicity tests that can be used for estimating
starting doses for acute oral toxicity tests, thereby reducing the number of animals used.

Related Activities by Foreign Governments

A new European Union Regulation on Registration, Evaluation, Authorization, and
Restriction of Chemicals (REACH) went into effect June 1, 2007. The main goals of REACH
are (1) to improve the protection of human health and the environment from risks associated with
chemicals in commerce and (2) to promote alternative test methods. REACH requires
manufacturers and importers to demonstrate they have appropriately identified and managed the
risks of substances produced or imported in quantities of one ton or more per year per company.
The new European Chemicals Agency  (ECHA)24 will manage the system databases, coordinate
evaluation of chemicals, and run a public database of hazard information25.

The European Centre for the Validation of Alternative Methods (ECVAM)26 coordinates
the validation of alternative test methods in the European Union. ECVAM develops, maintains,
and manages a database on alternative  procedures and promotes the development, validation, and
international recognition of alternative  test methods.

The Japanese Center for the Validation of Alternative Methods (JaCVAM) is part of the
Japanese National Institute of Health Sciences. JaCVAM has conducted validation studies for
alternative test methods and participates in international validation efforts27.

The Korean Center for the Validation of Alternative Methods (KoCVAM) is a branch of
NITR, the National Institute of Toxicological Research. NITR is collaborating with the Korean
Society for Alternatives to Animal Experiments (KSAAE) to refine methods in acute oral,
reproductive/development, genetic, and endocrine toxicity testing28.

The Organization for Economic Co-Operation and Development (OECD) represents 30
countries in the Americas (including the United States), Europe, and Asia. The OECD
"Guidelines for the Testing of Chemicals" provides a collection of internationally harmonized
testing methods for a number of toxicological endpoints using in vivo, in vitro, and even
alternative approaches.29 Test guidelines can be updated to reflect scientific advances and the
state of the science if member countries agree to do so. A few OECD workgroups and efforts
address issues relevant to this EPA strategy, e.g., the OECD QSAR Toolbox30 and the joint
OECD/IPCS (International Programme for Chemical Safety) Toxicogenomics Working Group,
which has developed a proposal for a Molecular Screening Project, modeled after EPA's
ToxCast™ program.
23 http://iccvam.niehs.nih.gov/methods/acutetox/inv nrutmer.htm
24 http://echa.europa.eu/reach en.asp
25 http://ec.europa.eu/environment/chemicals/reach/reach intro.htm
26 http://ecvam.jrc.it/
27 http://www.nihs.go.jp/english/index.html
28 http://wwwsoc.nii.ac.jp/isaae/PARK.pdf
29 http://titania.sourceoecd.org/vl=856000/cl=23/nw=l/rpsv/periodical/pl5 about.htm?jnlissn=1607310x
30 http://www.oecd.Org/document/23/0.3343.en 2649 37465 33957015  1 1 1 37465.00.html
                                           25

-------
Academia

Numerous U.S. academic researchers and centers are funded by NIH or EPA's National Center
for Environmental Research to develop assays and analysis methods that might be helpful to the
goals of this EPA research strategy. This includes two Bioinformatics Centers funded by EPA in
2006.

The European Commission funds several large academic, government, and industry consortia
that are conducting research that could lead to effective in vitro toxicity tests. The CASCADE
Network of Excellence31 studies human health effects of chemical residues and contaminants in
food and drinking water, designing assays to elucidate estrogen, testosterone, and thyroid
hormone pathways for the development of mechanism- and disease-based test methods. The aim
of the carcinoGENOMICS32 project is to develop in vitro methods for assessing the
carcinogenic potential of compounds. ReProTect33 is optimizing an integrated set of
reproductive/developmental tests for a detailed understanding of gametogenesis,  steroidogenesis,
and embryogenesis that can support regulatory decisions.

Industry

The European Partnership for Alternative Approaches to Animal Testing (EPAA)34 is a
joint initiative from the European Commission and a number of companies and trade federations.
Its purpose is to promote the development of alternative approaches to safety testing. The EPAA
focuses on mapping existing research; developing new alternative approaches and strategies;  and
promoting communication, education, validation, and acceptance of alternative approaches.

Non-Governmental Organizations (NGOs)

The Comparative Toxicogenomics Database35 (CTD) elucidates molecular mechanisms by
which environmental chemicals affect human disease. CTD includes manually curated data
describing cross-species chemical-gene/protein interactions and chemical- and gene-disease
relationships to illuminate molecular mechanisms underlying variable susceptibility and
environmentally influenced diseases. These data will also provide insights into complex
chemical-gene and protein interaction networks.

The Johns Hopkins Center for Alternatives to Animal Testing36 supports the creation,
development, validation, and use of alternatives to animals in research, product safety testing,
and education. Similarly, AltTox.org37 provides information on non-animal  methods for toxicity
testing including a table38 that summarizes the alternative testing methods by endpoint that have
been approved or endorsed internationally by at least one regulatory agency.
31 http ://www. cascadenet. org/
32 http://www.carcinogenomics.eu/
33 http://www.reprotect.eu/
34 http://ec.europa.eu/enterprise/epaa/index en.htm
35 http://ctd.mdibl.org/
36 http://altweb.jhsph.edu/index.htm
37 http://www.alttox.org/about/
38 http://www.alttox.org/ttrc/validation-ra/validated-ra-methods.html
                                           26

-------
REFERENCES

Coecke S, Goldberg AM, Allen S, Buzanska L, Calamandrei G, Crofton K, Hareng L, Hartung T, Knaut
H, Honegger P, Jacobs M, Lein P, Li A, Mundy W, Owen D, Schneider S, Silbergeld E, Reum T,
Trnovec T, Monnet-Tschudi F, Bal-Price A. (2007) Workgroup report: incorporating in vitro alternative
methods for developmental neurotoxicity into international hazard and risk assessment strategies. Environ
Health Perspect.  115(6):924-31.

Collins FS, Gray GM, Bucher JR. (2008) Transforming Environmental Health Protection. Science
319:906-7.

Dix DJ, Houck KA, Martin MT, Richard AM,  Setzer RW, Kavlock RJ. (2007) The ToxCast program for
prioritizing toxicity testing of environmental chemicals. Toxicol Sci. 95(1):5-12.

Edwards SW, Preston RJ.  Systems Biology and Mode of Action Based Risk Assessment. Toxicol. Sci. in
press, doi:10.1093/toxsci/kfnl90.

Guyton KZ, Kyle AD, Aubrecht J, Cogliano VJ, Eastmond DA, Jackson M, KeshavaN, Sandy MS,
Sonawane B, Zhang L, Waters MD,  Smith MT. (2008) Improving prediction of chemical carcinogenicity
by considering multiple mechanisms and applying toxicogenomic approaches. Mutat. Res.  In press,
doi:10.1016/j.mrrev.2008.10.001.

Judson R,  Richard A, Dix D, Houck K, Elloumi F, Martin M, Cathey T, Transue TR, Spencer R, Wolf M.
(2008) ACToR—Aggregated Computational Toxicology Resource. Toxicol ApplPharmacol. 233(1):7-13.

Kavlock RJ, Ankley G,  Blancato J, Breen M, Conolly R, Dix D, Houck K, Hubal E, Judson R,
Rabinowitz J, Richard A, Setzer RW, Shah I, Villeneuve D, Weber E. (2007) Computational Toxicology
A State of the Science Mini Review. Toxicol Sci.  Dec 7.

Kavlock RJ, Dix DJ, Houck KA, Judson RS, Martin MT, Richard AM. (2008). ToxCast™: Developing
predictive signatures for chemical toxicity. Proceedings of the 6th World Congress on Alternatives  in the
Life Sciences.T8-4-1. In Press.

Lein P, Locke P,  Goldberg A. (2007) Meeting  report: alternatives for developmental neurotoxicity testing.
Environ Health Perspect.  115(5):764-8.

National Research Council of the National Academies (NRC). (2006) Toxicity Testing for Assessment of
Environmental Agents. The National Academies Press. Washington, DC.

National Research Council of the National Academies (NRC). (2007) Toxicity Testing in the 21st
Century: A Vision and A Strategy. The National Academies Press. Washington, DC.

Shi L et al. (2006) The Micro Array Quality Control (MAQC) project shows inter- and intraplatform
reproducibility of gene expression measurements. Nat Biotechnol. 24(9):1151-61.

U.S. Environmental Protection Agency. (1995) Ecological risk:  a primer for risk managers. Washington,
DC:  U.S. Environmental Protection Agency. EPA/734/R-95/001. Available at http://www.epa.gov/nscep/.

U.S. Environmental Protection Agency. (1998) Guidelines for Ecological Risk Assessment. Washington,
DC:  U.S. Environmental Protection Agency. EPA/630/R-95/002F.
http://cfbub.epa.gov/ncea/cfm/recordisplay.cfm?deid=12460
                                              27

-------